A non-targeted metabolomics approach to quantifying differences in root storage between fast- and slow-growing plants


Author for correspondence:

Mark Rees

Tel: +44 114 222 0117

Email: m.rees@sheffield.ac.uk


  • Life history theory posits that slower-growing species should invest proportionally more resources to storage, structural (e.g. stems) or defence traits than fast-growing species. Previously, we showed that the slower-growing monocarpic plants had lower mortality rates and higher bolting probabilities after two defoliation events. Here, we consider a mechanistic explanation, that the slower-growing species invested relatively more resources to storage.
  • We compared the relative levels of root storage compounds between eight monocarpic species using metabolomic profiling, and characterized plant growth using a size-corrected estimate of relative growth rate (RGR).
  • Growth rate was negatively correlated with the proportional allocation of root metabolites identified as sucrose, raffinose and stachyose and with amino acids known for their roles in nitrogen storage, particularly proline and arginine. The total amount and concentration of energy-corrected carbohydrates were also negatively correlated with RGR.
  • Our results show for the first time that slower-growing species invest proportionally more of their total root metabolites in carbon- and nitrogen-storage compounds. We conclude that the increased investment in these reserves is an important resource allocation strategy underlying the growth–survival trade-off in plants.


Growth rate is seen as a key indicator of a species’ ecology and life history and is one of the most important axes of variation among species (Grime & Hunt, 1975; Madsen & Shine, 2000; Ricklefs & Wikelski, 2002). In plants, even when species are grown in isolation under favourable conditions, they display large variation in potential relative growth rates (RGR), with growth rate often varying by > 10 fold (Grime & Hunt, 1975; Poorter & Remkes, 1990). There are obvious benefits to a high growth rate, including an increased competitive ability for occupying space and capturing resources such as light and limiting soil nutrients (Poorter, 1989). If an individual can grow fast enough to monopolize these resources and subsequently reproduce, it should be more successful than slower-growing individuals. However, as there is large variation in growth rate between species, there must be important benefits to an inherently slow growth rate and costs of fast growth under certain conditions (Grime & Hunt, 1975; Chapin, 1980; Lambers & Dijkstra, 1987).

Slower-growing plants are generally found in low nutrient, unpredictable and ‘stressful’ environments. For example, slow-growing plants dominate when there is low available nitrogen or high salinity in the soil (Parsons, 1968; Grime & Hunt, 1975; Grime, 1977; Clarkson, 1985). Fast growers typically display more flexible responses to environmental conditions, responding to low soil nutrient availability by reducing photosynthetic rates and increasing their root capacity to absorb limiting nutrients (Chapin, 1980). However, fast-growing plants lack the physiological plasticity to survive in very low nutrient conditions. The low nutrient conditions that some plants are adapted to therefore place considerable restrictions on physiological plasticity, life history and ecological strategies. Nevertheless, it is thought that a low growth rate may not always be the result of a physiological limitation, but may actually be part of a broader landscape of functional trade-offs that increase survival in low nutrient conditions and other unfavourable environments (Ardent, 1997; Metcalf & Koons, 2007).

Allocation to traits that improve longevity and defence is especially important in slow growers, as life history theory assumes that the potential benefit of increased future fecundity by delaying reproduction is weighed against the chance of mortality in that period (McLaren, 1966; Koons et al., 2008). However, the wider picture of exactly how slow-growing species are able to live longer is incomplete. In particular, it is not well understood how growth rate interacts with allocation to long-term nutrient reserves, which could serve to buffer plants when environmental conditions are unsuitable to support growth, and aid re-growth after herbivory (Chapin et al., 1990).

Many species have taproots that are thought to be involved in storage (Steinlein et al., 1993). In plants, carbon, alongside nitrogen and phosphorus, is the most vital resource required for growth and maintenance. Plants store carbon as carbohydrates such as starch, sucrose and larger oligosaccharides, especially polysaccharides containing fructose molecules (Kandler & Hopf, 1982; Keller & Matile, 1985; Chapin et al., 1990; Keller & Pharr, 1996). However, other sources (e.g. lipids) (Beeson & Proebsting, 1988), can also be used for carbon storage. Plants are also known to store nitrogen and phosphorous in various ways (Chapin et al., 1990).

Currently, there is evidence that accumulated carbohydrate reserves are used by plants for respiration during dormancy (Kozlowski, 1992) and to replace shoot tissue after herbivory (Marquis et al., 1997). Reserves can be mobilized at the beginning of the growth season to support new tissue growth (Chapin et al., 1990; Kozlowski, 1992; Gaucher et al., 2005). In addition, in some species that are often exposed to fire, plants with high levels of stored starch are more likely to resprout, rather than regenerate from seed after fire damage, suggesting that some reserves are allocated as a back-up to aid regrowth after such an event (Pate et al., 1990; Bell et al., 1996). The idea that carbohydrates are used as buffers in many circumstances is therefore widely accepted, as is the general idea that there are life-history explanations for resource allocation differences between species. However, the present evidence for an explicit relationship between growth rate and carbohydrate storage is unconvincing.

Our understanding of the link between growth rate and carbohydrate storage comes primarily from trees. Several studies on trees have found that RGR was negatively correlated with sugar or nonstructural carbohydrate concentrations and total pool sizes (Myers & Kitajima, 2007; Poorter & Kitajima, 2007). This strategy increases the probability of survival in shady forests, while saplings are small. However, in the main, investigations that look for trade-offs between growth and storage in plants provide conflicting conclusions (Poorter & Bergkotte, 1992; Van Der Meijden et al., 2000; Metcalf et al., 2006; Poorter & Kitajima, 2007) and there has been very little work on short-lived plants.

Uncovering a trade-off between growth and storage has proven difficult for two principal reasons. First, the widespread use of the classical method for measuring RGR, which does not account for the size-dependence of growth, can result in trade-offs being masked by size effects (Turnbull et al., 2008). For example, trade-offs between growth and defence have proven difficult to detect, but new size-corrected analyses of RGR have recently provided good evidence for this trade-off in Arabidopsis thaliana (Paul-Victor et al., 2010). In order to remove the effects of size, we estimated RGR at a common plant size (Rose et al., 2009). Second, the methods used to measure storage are often very selective for specific families of compounds or use easy-to-measure proxies for storage. Therefore, it is likely that important information contained in a full metabolite profile is lost. For example, some studies have used dry root weight instead of, or as a proxy for, storage, although root biomass may not be an accurate indicator of readily mobilized carbon stores (Metcalf et al., 2006). Commonly, nonstructural carbohydrates are extracted from tissues, the complex carbohydrates hydrolysed to glucose, and total glucose is then measured (Mooney et al., 1995; Myers & Kitajima, 2007; Poorter & Kitajima, 2007). An alternative approach would be to use nontargeted mass spectrometry to detect many hundreds of metabolites in a tissue sample. The result is then a snapshot at a particular point in time of the metabolite pool contained in a plant sample, and presents a more complete picture of the compounds that are present and their relative abundances. Metabolomics, therefore, offers an insight into how individual compounds covary with other families of compounds and can also reveal relationships between particular metabolites and other measured plant traits. This metabolite profiling provides clues to whole-plant responses to stress or applied treatments (Bundy et al., 2009). The application of metabolomics in life-history research is particularly appropriate because it is possible to infer resource allocation priorities within and between organs. Significantly, if expected compounds are not found to change as predicted, the complete metabolomics data could reveal which other compounds are changing.

Monocarpic perennials typically have size-dependent reproduction, although in some species there is an additional weak age-dependent component (Rose et al., 2002). This is a strategy that spreads the risk of reproduction over time in a variable environment (Venable & Brown, 1988; Rees, 1994). Delayed reproduction can evolve in semelparous species when this results in higher fecundity (McLaren, 1966); however, the potential benefit of increased future fecundity by delaying reproduction is weighed against the chance of mortality in that period (Koons et al., 2008). Therefore, taproot reserves, which can be used for both reproduction and ‘risk-aversion’ (e.g. herbivory), may decrease the risk of mortality associated with delayed reproduction (Chapin et al., 1990). In addition, belowground storage is less vulnerable to herbivory than aboveground stores, which is suggestive of a significant role in maintenance and survival compared with stores in other plant organs. The taproot stores that are available for maintenance may ultimately be used for reproduction, and indeed previously we found a growth–reproduction trade-off in monocarpic perennials as well as a growth–survival trade-off after full defoliation, whereas in the absence of defoliation faster growers had an increased probability of flowering in the second year (Rose et al., 2009).

In this investigation, we use monocarpic plants to look at the relationships between growth rate and storage, focusing on the enlarged taproots observed in the chosen species. Monocarpic plants are ideal for studying allocation strategies, as their simple life history, with its single reproductive event, means that the complications involved with understanding the costs of reproduction, and how these influence returns from current and future reproduction, do not arise. Our work builds on a previous study that linked a lower size-corrected RGR in individual plants to higher survival and bolting probability after two full defoliation events (Rose et al., 2009), and uses a novel application of metabolomic methodology to analyse the relative amounts of carbohydrate reserves in roots. In order to look at the relationship between growth and storage, we used a nontargeted approach to identify the compounds that varied in the roots, followed by a targeted approach to analyse changes in more detail. Plants may vary their investment in storage in three ways: either by changing the relative abundance of different storage compounds; by changing the total concentration of compounds; or by altering the size of taproot, that is, the total pool size. As total pool size can be confounded with plant size, for example a large plant might have a larger total store despite having a lower allocation to storage, we measured the concentration and the relative abundance of carbohydrates. The metabolomics approach also enabled us to identify all compounds that are associated with differences in growth rate between species. After performing this non-targeted analysis, we then ask which of the compounds associated with growth are known storage compounds. By doing so, we provide a unique insight into the mechanisms of storage.

Materials and Methods

Plant material

The experiment took place at Tapton Experimental Gardens, University of Sheffield, UK. Seeds of Arctium minus L., Carduus nutans L., Cirsium vulgare Savi, Digitalis purpurea L. Senecio jacobaea L., Verbascum blattaria L. and Verbascum thapsus L. were sown between 15 and 21 March 2007 into degradable pots 7 mm in diameter and put into a glasshouse. The pots were filled with a 9 : 1 : 1 mixture of sand : vermiculite : M3 compost. After a few weeks of growth the plants were transferred into 2.2 l pots, filled with the same sand, vermiculite and compost ratio as before, and placed outside in a randomized, eight-block design, roughly balanced by species. Plant size measurements were taken weekly via the longest leaf length, as this is a good predictor of plant biomass in rosette-forming species (Metcalf et al., 2006). Individuals of the seven monocarpic species (= 150 in total) were harvested on the following dates: 12 May 2007 (H1) after c. 12 wk of growth, = 26; 22 June 2007 (H2) after c. 13 wk of growth; = 22, 7–8 July 2007 (H3) after c. 15 wk of growth = 27; 21–22 July 2007 (H4) after c. 17 wk of growth, = 23; 31 August 2007 (H5) after c. 23 wk of growth, = 22; and 5 March 2008 (H6) after c. 11.5 months of growth, = 30. During harvesting, individual root samples weighing 200–500 mg were taken from all plants. Each sample was a complete cross-section of the taproot and was taken from halfway down that main (or central) taproot. The samples were then flash-frozen in liquid nitrogen and transferred to a −80°C freezer until later analysis by mass spectrometry.

Identification of carbon-storing compounds in the root

We initially tested for the presence of starch in the root samples, as this is a widespread storage compound that is not easily extracted. We sectioned eight plants, two of each species, producing 60-μm thick cross-sections. After iodine staining, the sections were examined under a light microscope from low to high magnification. A wheat (Triticum aestivum L.) seed section was also iodine-stained as a positive control. The wheat endosperm contained starch granules of < 10 μm diameter in size.

Mass spectrometry to detect storage compounds and data formatting for further analysis

The root samples were prepared for mass spectrometry by tissue extraction using a (5 : 2 : 2) methanol : chloroform : water solution adapted slightly from the method described in Davey et al. (2008) (see Supporting Information Methods S1). Mass spectrometry is a method of compound detection within biological samples, by separating ions according to mass : charge (m/z) ratios. We used nontargeted metabolomics methods to analyse the relative amounts of soluble metabolites in roots because other methods such as gas chromatography and high-performance liquid chromatography are only selective for specific classes of compounds (Roessner et al., 2000; Rozan et al., 2001) and therefore lose much of the information that is present in a complete metabolite profile. The composition of compounds and their relative abundance is described as the ‘metabolome’, which allows simultaneous analysis of different groups of compounds and how they may covary in different species, environments or through experimental manipulations. The ion count for a given m/z value (the molecular size) corresponds to the relative concentration of the ion in the biological sample. The samples were analysed using electro-spray ionization with a triple quadrupole mass spectrometer, API Sciex III Plus, (AB Sciex UK Ltd, Warrington, UK), over a mass range of 100–1000 Da in positive mode. The data were peak-centred and three replicate runs for each sample were merged into one file with ion intensity averaged over the replicates into 1 Da mass bins. For OPLS (orthogonal partial least square analyses) each m/z bin represented a proportion of the total ion count (% total ion count) in the sample.

Confirmation of storage compound identification

The compounds of interest were pinpointed in two ways. First, a principal component analysis (PCA) was completed for several species in H5 to determine m/z bins that discriminated each species. In addition, mass lists for the root metabolomic profiles of all species were compared with the masses of known storage compounds. Second, m/z bins that were consistently associated with slow or fast growth in the OPLS analyses were identified.

Tandem mass spectrometry (MS/MS) was used to confirm the putative identification of amino acids. Fragmentation patterns of ion peaks of interest were compared with those of standards using a Q-Star QTOF mass spectrometer (AB Sciex UK Ltd). To provide a second method of confirming carbohydrate identity, gas chromatography was completed using a Zebron Inferno 2B-5HT Column, (Zebron, Macclesfield, UK).

We compared elution times from the column between standards and the peaks in the root sample preparations. As an internal standard, phenyl α-d-glucopyranoside was spiked to each sample and sugar standard.

Relative growth rate analysis

Growth rates were calculated as described by Rose et al. (2009). Growth curves for each individual in the present experiment could not be estimated accurately because the destructive harvesting limited the number of census dates. However, as the plants in the present experiment were part of a larger experiment, which was measured with more census times and for more individuals, = 842, we used species-average growth rates calculated using the data presented in Rose et al. (2009). Individual plant growth rates were then estimated at a common size on 19 June 2007. Previously, RGR on 19 June 2007 was found to predict subsequent survival and reproduction following defoliation (Rose et al., 2009).

RGR relationships with storage compounds

In the OPLS analysis the binned percentage ion count data for harvests one, five and six (H1, H5, H6) were analysed using the commercially available simca-p software, (Umetrics AB, Umeå, Sweden). Orthogonal partial least square analyses is an established method for the analysis of metabonomic and metabolomic datasets, with the aim of extracting variation in the x-matrix (in our case derived metabolite profiles) that is associated with one or more y-variables (species growth rates). For more information about the OPLS algorithm and a detailed explanation of this multivariate method, see Trygg et al. (2007) and Eriksson et al. (2006). The harvests were chosen because they represented different stages of plants’ lifespan and were also completed at different times of year: seedling (late spring), nonflowering rosette (late summer/autumn) and just before bolting occurred (spring the following year). The OPLS analyses were achieved with the simca-p autofit option, using the species-average size-corrected growth rate as the y-variable and the individual sample binned ion-intensity data (x-variable). The data were Pareto-scaled to allow peaks of all sizes to be given a more equal weighting (Noda, 2008). Alternative analyses were also performed averaging the ion intensity data for each species. This did not change the results in any substantive way, although the OPLS CV-ANOVA (Eriksson et al., 2008) tests became more significant, and so we only present the analyses based on the individual sample data.

Excluding the multivariate analyses, all statistics were completed using the R statistical package software (R Development Core Team, 2009). We used ANOVA with species-average growth rate, harvest (H1–H6) and interactions between them as explanatory variables. All response variables were log-transformed. Analyses suggested that sucrose and the raffinose family of oligosaccharides (RFOs) raffinose, stachyose and verbascose played an important role in root storage. The response variables were therefore the ion intensities corresponding to masses of sucrose, raffinose, verbascose, stachyose, total carbohydrates (sucrose and the RFOs) per 0.1 g of taproot and total carbohydrates in the taproot. To extract the mass intensity data we took the largest intensity within a 0.5 Da confidence interval of the sugar's monoisotopic mass plus adducts. Calculating total carbohydrates and correcting for the energy value of each compound was also possible as all the sugar compounds are constructed from galactose unit extensions to a sucrose base unit. We scaled total carbohydrate concentration to the taproot, using fresh-weight taproot measures. This provided a measure of total carbohydrates in the taproot. The full statistical models are presented as the model represented the structure of the experimental design.


Storage compounds

Starch granules could not be detected in any of the iodine-stained root sections (Fig. 1). The wheat endosperm (positive control) contained very small starch granules; thus, if present in the root sections, these would also have been detected. Therefore, no substantial amounts of starch were detected in any species. Monoisotopic masses (accounting for adducts) corresponding to sucrose, raffinose, stachyose and verbascose were identified as potential storage compounds from the PCA analysis (Fig. S1a,b) and root metabolomic profiles. The presence of these compounds was confirmed using gas chromatography (GC) analysis (Table S1). Within the small sample of extracts selected for GC analysis, all four compounds were found to be present in V. blattaria, A. minus, S. jacobaea, D. purpurea and C. nutans; sucrose, raffinose and stachyose were present in V. thapsus and sucrose and raffinose was present in C. vulgare.

Figure 1.

Images of iodine-stained root sections. (a) Arctium minus, (b) Cirsium vulgare (c) Carduus nutans, (d) Digitalis purpurea, (e) Senecio jacobaea, (f) Verbascum blattaria, (g) Verbascum thapsus, (h) wheat (Triticum aestivum L.) seed with stained starch granules. Larger, more densely packed starch granules are found towards the centre of the seed, while smaller granules are found at the edges.

The results of the OPLS analyses (Figs 2, 3, 4) indicated that some masses, likely to be amino acids, showed consistent association with variation in growth rate. In order to establish that amino acids were present in the root samples, the MS/MS peak profiles of three amino acid standards; proline, arginine and glutamine were compared with targeted ion peak profiles in each species. Comparison of the MS/MS profiles indicated that all the plant species had the three amino acids present in the roots (see Fig. S2a,b for an example of proline identification within a sample). There was evidence that there were other compound(s) present within some of the targeted ion peak (each targeted peak had a 1 Da range). However, based on relative fragmentation peak intensities, the amino acids constituted the significant majority of the ions within the targeted peaks.

Figure 2.

Orthogonal partial least-square analyses (OPLS) analysis of harvest 1 (H1). The graphs show the 25 m/z bins most positively (a) and negatively (b) associated with species mean relative growth rate (RGR) (as the y-variable). The graphs are extracted from the full analysis of 1000 m/z bins. The ‘X loading weight p’ (y-axis) is the predictive component which explains 5.95% of the variation. A negative sign indicates a negative association with RGR. The error bars were calculated by jack-knifing (Efron & Gong, 1983). The confidence level used for calculating these confidence intervals was 95%. R2X = 0.27, R2Y = 0.983, Q2 = 0.445, = 26. CV-ANOVA: F = 2.53, P = 0.057. These are defined as: R2X, the fraction of the variation of the X variables explained by the model; R2Y, the fraction of the variation of the Y variables explained by the model; Q2, the total variation of the X variables predicted by the model.

Figure 3.

Orthogonal partial least-square analyses (OPLS) analysis of harvest 5 (H5). The graphs show the 25 m/z bins most positively (a) and negatively (b) associated with species mean relative growth rate (RGR) (as the y-variable). The graphs are extracted from the full analysis of 1000 m/z bins. The ‘X loading weight p’ (y-axis) is the predictive component, which explains 17.8% of the variation. A negative sign indicates a negative association with RGR. The error bars were calculated by jack-knifing (Efron & Gong, 1983). The confidence level used for calculating these confidence intervals was 95%. R2X = 0.462, R2Y = 0.973, Q2 = 0.781, = 22. CV-ANOVA: F = 8.93, < 0.001. These are defined as: R2X, the fraction of the variation of the X variables explained by the model; R2Y, the fraction of the variation of the Y variables explained by the model; Q2, the total variation of the X variables predicted by the model.

Figure 4.

Orthogonal partial least-square analyses (OPLS) analysis of harvest 6. The graphs show the 25 m/z bins most positively (a) and negatively (b) associated with species mean relative growth rate (RGR) (as the y-variable). The graphs are extracted from the full analysis of 1000 m/z bins. The ‘X loading weight p’ (y-axis) is the predictive component, which explains 8.07% of the variation. A negative sign indicates a negative association with RGR. The error bars were calculated by jack-knifing (Efron & Gong, 1983). The confidence level used for calculating these confidence intervals was 95%. R2= 0.45, R2= 0.991, Q2 = 0.713, = 30. CV-ANOVA: F = 4.73, P < 0.01. These are defined as: R2X, the fraction of the variation of the X variables explained by the model; R2Y, the fraction of the variation of the Y variables explained by the model; Q2, the total variation of the X variables predicted by the model.

Relationship of species-mean RGR with root metabolites

Figs 2–4 show the OPLS analyses of the derived root metabolomic profiles and species-mean RGR (Table 1). The graphs in Figs 2–4 show the 25 m/z bins most negatively associated with growth rate, that is, slower growth, and the 25 m/z bins most positively associated with growth rate that is, faster growth. In addition, Fig. S3 may aid interpretation of these OPLS loading plots. Overall, there were 1000 m/z bins in the full loading plot, and so the chance of a m/z bin being consistently associated with fast or slow growth rate in the three harvests is exceptionally small. For example, storage compounds can occur in 12 bins (sucrose, raffinose, verbascose and stachyose, and each can occur in up to three bins) so the probability of observing a storage compound is 12/1000. If we draw a sample of 25 of these then the chance of observing three bins with storage compounds at a particular harvest is c. 0.003 (using the binomial distribution). For raffinose, which was identified as negatively associated with growth in all three harvests, the probability is c. 0.0003 (binomial distribution, P = 3/1000, n = 25; raised to the power of 3). These are conservative figures, as in several cases storage compounds were identified as the compound most negatively associated with growth. In all analyses, raffinose (mass 527) was one of the compounds most highly associated with slow growth. Sucrose (masses 365, 381) was highly negatively associated with growth rate in H1 (Fig. 2a) and H6 (Fig. 4a) and the oligosaccharide stachyose (masses 689, 705) in H1 (Fig. 2a). No strong positive associations were found between growth and sucrose or the RFOs (Figs 2–4).

Table 1. Species-mean RGR
SpeciesSpecies mean growth rate (mm mm–1 d–1) SE around species mean growth rate
  1. We used nonlinear mixed effects models to fit growth curves for individual plants (= 842), using the dataset and fitting methods described in (Rose et al., 2009). A size corrected average relative growth rate (RGR) estimate at 19 June 2007 was made for each species. This approach was necessary, as most of the plants taken for harvesting in this experiment did not have a sufficient number of census measurements to accurately fit individual growth curves. However, despite the high variation in individual RGR within a species, species mean RGR was still a good predictor of survivorship and bolting probability during the subsequent summer (M. Rees, unpublished). We calculated species mean RGR, based on the method outlined in Rose et al. (2009), from the fit of a 3-parameter von Bertalanffy growth curve given by Li(t) = L∞,i(1 − exp[− ki (t − t0,i)]) where Li(t) is the log size of the ith plant at time t, L,i is the asymptotic size, ki is the rate constant and t0,i the time at which Li(t) = 0.

Verbascum thapsus 0.022820270.0002200908
Digitalis purpurea 0.020777850.0002376544
Carduus nutans 0.019100120.0003648205
Verbascum blattaria 0.018026580.0002957425
Arctium minus 0.015869050.0006418032
Cirsium vulgare 0.015882780.0006136966
Senecio jacobaea 0.015804220.0002593089

Slow-growing species were also associated with increased allocation to masses identified as amino acids. For example, in H1, masses corresponding to proline (116) and arginine (175) were negatively associated with RGR (Fig. 2a). In H5, masses corresponding to proline (116) and asparagine (133) were associated with slow growth (Fig. 3a), and in H6, masses corresponding to proline (116), arginine (175), cysteine (122) and glutamine (148) were negatively associated with RGR (Fig. 4a). In all analyses, only one mass corresponding to an amino acid was associated with fast growth, which was tryptophan (205) in H6 (Fig. 6b). In H6 only, faster growers accumulated metabolites in the mass bin 205 (possibly mannitol and/or sorbitol, as well as/or tryptophan) (Fig. 4b), and in H5 and H6, 413 (possibly fucosterol or stimasterol, which are both phytosterols) (Figs 2b–4b). Fast growth was also associated with several unidentified m/z bins, such as 316, 413, 809, 502, 517 and 691.

Carbohydrate concentration

Taproot concentration of total carbohydrates was negatively correlated with species average growth rate at harvests 1 and 6 (Fig. 5, Table 2). This was mainly because of highly significant correlations between sucrose (Fig. S4) and raffinose concentrations (Table 2) and slow growth, as these two compounds comprised the majority of the root storage ions, with a mean across all plants of 77% of the total energy-corrected carbohydrate concentration. Stachyose and verbascose together in the taproots made up 23% of the total energy-corrected carbohydrate concentration. The relationship of carbohydrate concentration (sucrose and the RFOs) with growth rate did alter with harvest (Fig. 5, Table 2).

Figure 5.

The relationship between average species growth rate and relative taproot carbohydrate concentration. The data comprised mass spectrometer counts of carbohydrates (sucrose and the raffinose family of oligosaccharides (RFOs)) in the extraction of 0.1 g root material. Open circles represent raw data and each line represents the model fit through the data. Two data points in harvest 4 were an order of magnitude larger than the rest of the data (both above 23 on the y-axis) and not shown in the graph; one data point was from the fastest growth rate and one from the cluster of slow growth rates. However, these data were included in the ANOVA in Table 2.

Table 2. ANOVA analyses showing the relationships of species-mean RGR and harvest with root carbohydrate storage
 dfSucrose concentrationRaffinose concentrationCarbohydrate concentrationStachyose concentrationVerbascose concentrationCarbohydrates in whole root
  1. The response variables are sucrose, raffinose, verbascose, stachyose, combined carbohydrate concentration in 0.1 g of taproot and total carbohydrates in the taproot. We used ANOVA with species-average growth rate (mm mm−1 d−1), harvest and interactions between them as explanatory variables. All response variables were log-transformed. F-values and significance levels (*, < 0.05; **, < 0.01; ***, < 0.001) are given in the body of the table. RGR, relative growth rate.

Spp RGR1           17.39***           21.26***        6.81*       0.98       0.24           39.68***
Harvest5             8.00***           2.78*          3.86**       0.79       1.80             29.77***
Spp RGR × Harvest5         1.90         1.27         2.69*         2.75*         3.04*          2.38*
R 2           0.33         0.23         0.22         0.12         0.15         0.64
Model error          0.63         0.53         0.55         0.63         0.68         1.32
Residuals 138138138138138114

Total root carbohydrates

The relationship between growth rate and total carbohydrates in the root was highly significant (Table 2), with fast-growing species plants having significantly less total taproot carbohydrates (Fig. 6, Table 2). Over harvests, the difference in total taproot carbohydrates between slow and fast growers increased. The effect of growth rate differed marginally between harvests (Table 2).

Figure 6.

The relationship between species average growth rate and total carbohydrates in the taproot in each harvest. Open circles represent raw data and the smooth line in each graph is the model fit, presented in Table 2. The data is made up of mass spectrometer counts of carbohydrates (sucrose and the raffinose family of oligosaccharides (RFOs)) in the extraction of 0.1 g of taproot biomass, which have been scaled up to the taproot fresh weight for each individual's plant root.


Our findings showed that slower-growing monocarps invested proportionally more of their root carbon resources than fast-growing species in specific carbohydrates and amino acids, particularly sucrose, raffinose and proline. These compounds were also present at greater relative concentrations and absolute amounts in slow-growing than fast-growing species. When scaled to the total taproot in order to account for interspecific variation in root size, slower-growing species also had a larger total carbohydrate pool. The findings demonstrate how differences in the type, concentration and absolute amount of carbon were linked with growth. Furthermore, our results showed a strong negative relationship between nitrogen-storing amino acids, such as proline and arginine, and species growth rate. The relationships between growth rate and these nutrient-storing compounds suggests that long-term reserves of both carbohydrates and nitrogen are critical storage traits. Our approach provides a novel insight into resource allocations to the root organ. We have shown the significance of carbon and nitrogen stores in slow-growing species by demonstrating that specific carbohydrates and amino acids are among the compounds that fluctuate most between species of varying average growth rates. This in turn provides strong evidence that the roots of the slow-growing species had a key role in resource storage, as variation in root size or total carbohydrate pools alone do not provide sufficient evidence for this.

We found that slower-growing species accumulated high levels of sucrose and raffinose when m/z bins were expressed as a percentage of the total ion count in the root. These m/z bins were also within the top 5% out of a 1000 m/z bins that could have been associated with low RGR (Figs 2, 3, 4). This is a strong indication that allocation to sugars is important for the slower-growing species. Sugars and the raffinose series of polysaccharides are known nonstructural carbon storage components that accumulate in plant tissue and can be easily mobilized by the cell (Chapin et al., 1990). As these sugars are osmotically active, they may also have significant roles in variable and seasonal environments. For example, they may be involved in frost tolerance (Bloom et al., 1985). The link between proportional allocation to root carbon-storing carbohydrates and interspecific variation in growth rate provides a mechanism to explain the growth-survival trade-off and our previous results, which demonstrated the benefits of slower growth to both survival and future reproduction following defoliation (Rose et al., 2009).

Slow-growing species not only allocated proportionally more of their total detected metabolites to sucrose and raffinose storage than fast growers, they also accumulated these compounds at higher concentrations (Table 2). After accounting for sucrose and the RFOs together, growth rate continued to have a significant negative relationship with carbohydrate concentration, although the effect was more dependent upon harvest date (Table 2, Fig. 5). This result was probably due to the presence of the larger oligosaccharides, stachyose and verbascose as, individually, the relationships of these compounds with relative growth rate were weak and dependent upon harvest (Table 2). Figs 2–4 also show that the relative allocation to stachyose and/or verbascose was only highly negatively associated with RGR in one harvest (stachyose in harvest 1) out of the three harvests analysed. Finally, when scaled to the total FW of each individual, the relationship between growth rate and combined root carbohydrates was highly significantly negative (Table 2, Fig. 6). This relationship remained consistent over all six harvests, from 12 wk after germination in spring 2007, until spring the following year, c. 2 months before flowering occurred. The consistency in the compounds’ relationships between harvests suggests that despite possible short-term dynamic changes in specific storage compounds, that there are fundamental resource allocation differences between the species, and these were not an artefact of differential responses to a short-term environmental change.

Ions identified as proline, arginine and other amino acids were associated with slower growth rate (Figs 2–4). In plants, amino acids are known to be important nitrogen reserves, with proline and arginine being the most significant (Sagisaka & Araki, 1983; Chapin et al., 1986; Sagisaka, 1987; Ohlson et al., 1995). In internal transport networks between plant organs, amino acids are the most common nitrogen carriers (Okumoto & Pilot, 2011). Amino acids are also synthesized into proteins, which are more complex nitrogen storage structures. Nitrogen limitation in natural environments is extremely common, therefore an accumulation of amino acids ensures that growth can be supported when this occurs (Chapin et al., 1990; Pate et al., 1990; Bell et al., 1996). Significantly, the strong relationship between the amino acids, as well as the carbohydrates, and a slow growth rate remained consistent from seedling to pre-reproductive adulthood.

Our results therefore provide clear evidence that investment in nitrogen root reserves is prioritized in slower growers compared with faster growers. Our findings may seem inconsistent with those of Poorter & Bergkotte (1992) and Poorter et al. (1990), who found that higher nitrate concentration in the whole plant, and the root in the first study, was positively associated with growth rate. However, we did not measure total nitrates directly, but found a relationship between several amino acids that are known for nitrogen storage. In this way, we have revealed information about how nitrogen is stored within the root organ, which has the potential to uncover further details regarding whole-plant functioning, in contrast to measures of total nitrogen or carbon. For example, amino acids also have other important physiological roles in the plant; they may be important osmolytes. In addition, the soil composition we made up was purposefully a low-nutrient mix, which could have altered allocation strategies in comparison with previous findings. By using a low-nutrient mix, we aimed to demonstrate the possible benefits of slow growth in a low-nutrient environment, and to find a possible mechanistic explanation for these benefits (see Rose et al., 2009), given that the advantages of fast growth are more easily revealed.

The metabolomics method allowed us to look at the relationships between growth rate and specific compounds without pre-judging which compounds to select. For example, this method identified that proline was consistently associated with slow growth. Previous work shows that proline is not only known for nitrogen storage but also accumulates under physiological stress in many plants. This amino acid may also be important for cold tolerance (Wanner & Junttila, 1999), drought stress (e.g. Zrust, 1994) and salt tolerance (Stewart & Lee, 1974). Mechanistic explanations for how proline may be involved in stress tolerance are discussed by Hare & Cress (1997). The relatively high allocation to this compound in the root is consistent with the assumption that slow growers invest more in maintenance traits and with ecological and life-history theory.

In the OPLS analyses, the only amino acid positively associated with fast growth was putatively identified as tryptophan. Tryptophan, which is involved in growth and plant developmental regulation, is a precursor for the production of auxin and defence chemicals and is used in protein synthesis (Radwanski & Last, 1995). Tryptophan could be a more common amino acid for nitrogen storage in some or all of the faster-growing species. However, the strong positive association (within the top 5% of m/z bins) with growth rate was not consistent through the three harvests. Also highly associated with fast growth was the m/z bin 413 (in two harvests), which was thought to be a phytosterol and the m/z bin 205, which was also putatively identified as mannitol and/or sorbitol (H6 only). Phytosterols such as stigmasterol regulate membrane and cell processes (reviewed in Piironen et al., 2000), while mannitol and sorbitol are sugar alcohols, which are an important carbon transporter in plants (Bloom et al., 1985). Again, the correlation between the putatively identified sugar alcohols and high RGR was not consistent over harvests, contrasting with low RGR being highly associated with larger carbohydrates in all three harvests. Therefore, the presence of sugar alcohols and tryptophan in the roots of faster-growing species could indicate a more short-term nutrient store.

Our results provide a mechanistic explanation for variation in growth rate in a group of plants where we know the effects of defoliation on subsequent survival and reproduction are linked with the growth strategy (Rose et al., 2009). The nontargeted metabolomics approach allowed us to quantify the abundance of nutrient-storage compounds in the context of the entire root metabolite fingerprint, thus providing a unique insight into the mechanisms of storage. We have demonstrated that there is a broad scope for metabolomic approaches in novel contexts.


We thank Heather Walker and Bob Keen for their help with mass spectrometry and gas chromatography. R.R.L.A. is grateful for support from a NERC studentship.